Urban Currents

The influence of spatial and social context on cyclists’ anticipatory yielding behavior

Two controlled video simulation experiments with 67 participants analyzed how cyclists make anticipatory yielding decisions when encountering crossing pedestrians. By measuring keyboard responses (Continue or Brake) to first-person traffic videos, the study tested the effects of kinematics, pedestrian density, and the presence of dedicated cycling lanes. Participants braked less frequently and responded slower in low pedestrian-density environments, when riding in dedicated cycling lanes, and when presented with larger time-to-arrival and distance gaps.

Why it matters — The study isolates how specific spatial and social design factors independently alter cyclist braking decisions, demonstrating that dedicated cycling infrastructure and lower pedestrian densities reduce a cyclist's propensity to yield.

Caveat: The findings are based on a small-sample laboratory video simulation rather than real-world cycling behavior.

Transportation Research Interdisciplinary Perspectives · pdf · doi

Large language models for travel behavior prediction

This study evaluates three large language model (LLM) frameworks—zero-shot direct prompting, textual-gradient prompt optimization, and supervised prediction using text embeddings—for predicting travel behavior. The frameworks were benchmarked against multinomial logit, random forests, neural networks, and TabPFN using Swissmetro mode choice, London mode choice, and NHTS trip-purpose datasets under varying labeled-data budgets. The results show that LLMs outperform or match traditional tabular baselines in data-scarce environments, and diagnostic tests confirm they respond to underlying travel-time and cost structures rather than relying on memorized benchmarks.

Why it matters — It establishes that LLMs can serve as viable, competitive predictors for travel behavior when labeled data is highly scarce, bypassing the need for extensive manual feature engineering or expert-crafted prompts.

Caveat: The study notes that as the volume of labeled training data increases, conventional supervised and tabular models eventually outperform the LLM-based approaches.

Transportation Research Interdisciplinary Perspectives · doi

FROM JUST ‐ IN ‐ TIME TO JUST ‐ IN ‐ CASE: Supply Chain Urbanism in San Diego‐Tijuana

This study analyzes the rapid growth of the logistics sector in the San Diego-Tijuana cross-border region, comparing the stagnant period of 2003-2013 with the unprecedented expansion of employment and commercial traffic from 2013 to 2023. It identifies trade law changes, transpacific supply chain disruptions, and transnational infrastructure investments as the primary drivers of this development.

Why it matters — The research documents a fundamental shift in regional logistics strategy from a 'just-in-time' efficiency model to a 'just-in-case' warehousing model, demonstrating how global supply chain vulnerabilities physically reshape cross-border urban landscapes.

Caveat: The analysis is restricted to a single, highly integrated cross-border metropolitan region, which may limit its applicability to non-border logistics hubs.

International Journal of Urban and Regional Research · doi

Altered by Gold: A Comparative Zooarchaeological Analysis of 19th to 20th Century Chinese and British Migrants in Ravenswood, a Northern Queensland Gold Mining Town

This study conducts a comparative zooarchaeological analysis of late 19th to early 20th century food waste from British and southern Chinese migrant groups in the gold mining town of Ravenswood, Queensland. The analysis reveals that the Chinese diet was diverse, featuring pork, poultry, and fish, while the British diet was dominated by mutton. It also documents instances of dietary crossover, including kangaroo and mutton consumption in a Chinese context and heavy pork consumption in a British context.

Why it matters — The findings demonstrate that migrant food choices in colonial frontier settings were dynamic and experimental rather than strictly determined by ethnic background, showing how groups adapted to local resource availability to forge new local identities.

Caveat: The study relies on archaeological food waste from a single mining settlement, which may limit how far these dietary patterns can be generalized to other colonial or urban contexts.

Archaeology in Oceania/Archæology & physical anthropology in Oceania · doi

Understanding urban heat patterns in Dhaka city using explainable GeoAI and satellite-derived LST

The study developed an explainable Geospatial Artificial Intelligence (GeoAI) framework using a LightGBM regressor to model summer land surface temperatures (LST) across Dhaka, Bangladesh, during May–August 2020 at a 500 m grid resolution. The model achieved high predictive accuracy (R² = 0.725, RMSE = 1.18 °C) and utilized GeoShapley to decompose predictions into non-spatial feature contributions, intrinsic location effects, and location-feature interactions. The analysis identified built-up intensity as the primary driver of elevated LST, while mapping localized cooling thresholds for vegetation and water bodies.

Why it matters — It introduces a method to isolate spatial context and localized interaction effects from raw physical features, allowing planners to identify specific geographic areas where green and blue infrastructure will yield the most effective cooling benefits.

Caveat: The model relies on land surface temperature as a proxy for ambient urban heat, which may not fully capture the actual thermal comfort experienced by residents at ground level.

Urban Informatics · pdf · doi

Analysis of urban green space distribution patterns and associated environmental externalities in Azamgarh city, India

This study mapped and analyzed the spatial-temporal distribution of urban green spaces and land surface temperatures in Azamgarh, India, using 30-meter resolution Landsat satellite imagery from 2000, 2011, and 2022. The analysis tracked changes in land use, Normalized Difference Vegetation Index (NDVI), and Land Surface Temperature (LST) to calculate per capita and per-unit area green space availability. The findings show a continuous decline in green cover over the 22-year period, which directly correlates with rising local surface temperatures and an expansion of the city's high-temperature zones.

Why it matters — It provides empirical evidence of how rapid conversion of vegetation to impervious surfaces accelerates the urban heat island effect in a rapidly growing, medium-sized Indian city. This establishes a direct, localized link between declining per capita green space and rising thermal stress to inform local urban planning interventions.

Caveat: The study relies on 30-meter resolution satellite data, which may not capture smaller, fragmented patches of urban vegetation that contribute to local cooling.

Discover Cities · pdf · doi

An iterative framework to inform urban design with computer vision

This study presents the Detect-Diagnose-Predict-Design (D2PD) framework, a conceptual workflow designed to integrate computer vision analytics directly into urban planning. The framework was developed through a systematic literature review of 27 selected papers from an initial pool of 500, followed by a reflexive thematic analysis.

Why it matters — It bridges the gap between real-time crowd monitoring and physical urban intervention, transforming computer vision from a passive surveillance tool on operational dashboards into an active, evidence-based generator of design guidelines.

Caveat: The proposed framework is conceptual and its practical utility in real-world planning workflows has not yet been empirically tested.

Discover Cities · pdf · doi

Framing social innovation in EU urban policy: Insights from the urban innovative actions (UIA) community initiative

This study analyzes how the European Union's Urban Innovative Actions (UIA) initiative (2014–2020) frames and regulates social innovation as an urban governance tool. Using critical policy analysis and an embedded case study of the ToNite project in Turin, Italy, it examines how the program's emphasis on place-based experimentation, multi-actor partnerships, and scalability shapes local practices.

Why it matters — It demonstrates that while EU-led funding frameworks enable experimental urban interventions, they simultaneously restrict the scope of grassroots social innovation by imposing selective criteria and rigid performance-driven logics.

Caveat: The empirical findings rely primarily on a single case study of the ToNite project in Turin to illustrate these policy dynamics.

Cities · doi

Spotting rat-runs: A network-based method for identifying potential cut-through routes in Bogotá, Colombia

The researchers developed a network-based screening method to identify local streets vulnerable to cut-through traffic, or rat-running, during peak congestion. They applied the model to Bogotá, Colombia, using OpenStreetMap data, observed speeds on major corridors, and traffic offense reports to calculate changes in travel-time-weighted betweenness centrality (ΔBC) between free-flow and congested states. A multilevel logistic model confirmed that residential streets with higher ΔBC had a statistically significant increase in the probability of traffic violations.

Why it matters — It provides a proactive, city-scale screening tool that relies on open data rather than expensive, hard-to-obtain historical traffic counts, allowing planners to target traffic-calming interventions and enforcement resources to vulnerable neighborhoods.

Caveat: The study relies on traffic offense reports as a proxy for actual traffic presence, which may reflect local enforcement intensity rather than the true volume of cut-through vehicles.

Computers Environment and Urban Systems · doi

Unveiling spatiotemporal mechanisms of urban traffic: Multi-scale determinants and explainable street-level dynamics of a graph neural network in Berlin

Researchers developed a two-stage framework combining a GraphSAGE neural network with Geographical and Temporal Weighted Regression (GTWR) to predict and interpret street-level traffic speeds across Berlin, Germany. Using Uber Movement speed data as the target variable alongside OpenStreetMap road attributes, population density, and street-view-derived activity indicators, the model predicted speeds for the 72% of the road network lacking empirical data with a mean absolute error of 5.73 km/h. The GTWR post-hoc analysis identified four spatial clusters of feature importance, showing that road type and speed limits increased predicted speeds while population density, pedestrian counts, and traffic signals decreased them.

Why it matters — The framework provides a method to interpret the black-box predictions of graph neural networks in urban transport, mapping how the model shifts its reliance on different spatial features across different neighborhoods. This spatial clustering of model behavior offers a structured pathway for transferring traffic prediction models to other cities where empirical speed data is missing.

Caveat: The model's interpretation and predictions rely on Uber Movement data as the ground truth, which may contain inherent sampling biases toward specific times or driver demographics.

Journal of Transport Geography · doi

Operationalizing accessibility thresholds: Identifying perceived minimum and sufficient accessibility in Canada

Using data from 15,960 participants in the Canadian Mobility Survey, this study establishes two perception-based accessibility thresholds—Perceived Minimum Accessibility (PMA) for basic access and Perceived Sufficient Accessibility (PSA) for convenient access—to grocery stores, healthcare, and leisure destinations via walking and public transit. The researchers combined binned log-linear regressions with a quartile-based approach to calculate these thresholds across different Canadian regions, mapping their spatial distribution and modeling their impact on weekly transit and walking mode shares.

Why it matters — It provides empirical, perception-based benchmarks for transport planning rather than relying on arbitrary theoretical thresholds, demonstrating that while basic minimum access (PMA) is widely met in large regions, only the higher convenience threshold (PSA) is associated with a substantial shift toward sustainable travel modes.

Caveat: The calculated accessibility metrics are compared against self-reported perceptions of accessibility, which may introduce subjective bias based on individual expectations and local environments.

Journal of Transport Geography · doi

Navigating housing system complexity – Part II: Quantitative analysis of the Housing System Abstraction Hierarchy (HSAH)

This study quantifies the Housing System Abstraction Hierarchy (HSAH) framework using network analysis metrics on a system model containing 176 nodes and 7,822 interconnected links. The analysis identifies that economic, environmental, and technical nodes have the highest structural influence via Eigenvector Centrality, while social and governance nodes act as critical bridges with high Betweenness Centrality. These structural relationships are applied to analyze cascading vulnerabilities and policy pathways for achieving adequate housing in Iran.

Why it matters — It translates a complex, qualitative conceptual framework of housing systems into a quantifiable network model. This allows policymakers to mathematically identify which specific sectors act as systemic leverage points or structural bottlenecks rather than relying on intuitive planning.

Caveat: The quantitative network structure and its resulting policy insights are demonstrated using a single national case study of Iran, which may limit direct transferability to countries with vastly different governance or economic structures.

Journal of Urban Management · doi

“Crossing the river by feeling the stones”: Sub-national agency in China's highway maintenance marketisation

This study analyzes how highway maintenance marketization is implemented in two underdeveloped Chinese provinces using document reviews and interviews with sub-national officials and practitioners. It identifies four narrative frames used by policymakers to justify reforms and documents how New Public Management principles are selectively adapted into hybrid arrangements that combine bureaucratic control with quasi-market practices dominated by state-owned enterprises.

Why it matters — It demonstrates that sub-national agency and deliberate inaction serve as pragmatic coping mechanisms when central policy expectations conflict with local socio-economic realities, challenging the assumption that Western public management reforms are directly transplanted into the Chinese administrative context.

Caveat: The findings are based on qualitative data from two underdeveloped provinces and may not represent the dynamics of highway maintenance reforms in wealthier Chinese regions.

Journal of Urban Management · doi

Low-carbon transition at community-level: A guiding evaluation framework and empirical investigations from Yangtze River Delta Region of China

This study develops a community-level low-carbon evaluation framework that integrates hardware dimensions like building energy efficiency and technology adoption with software dimensions such as residential behaviors and governance. The framework incorporates carbon sources from waste, transport, and energy alongside ecological sinks like green spaces, utilizing an indicator checklist and a four-step iterative evaluation process. Its application is demonstrated through an empirical investigation of a single case-study urban community in China's Yangtze River Delta Region.

Why it matters — It provides a structured diagnostic toolkit that allows local planners to measure both physical infrastructure and social behaviors simultaneously, identifying specific community-level strengths and weaknesses to guide targeted decarbonization strategies.

Caveat: The empirical validation of the framework is limited to a single case-study community in one region of China, which may limit immediate generalizability to different national or institutional contexts.

Journal of Urban Management · pdf · doi

Modeling spatial patterns of urban-rural landscape using landscape metrics and ecological quality index

The study analyzed land use patterns and ecological quality in the Greater Isfahan region, Iran, by classifying Landsat 9 imagery into six land cover types using a Support Vector Machine with 94% accuracy. Landscape fragmentation and connectivity were measured across eight geographical transects divided into 3 by 3 km blocks using five landscape metrics, while a direction-specific Ecological Quality Index (EQI) was constructed using Principal Component Analysis. The analysis revealed that the mean patch size of impervious surfaces shrinks from over 28 ha in the center to under 1 ha at the periphery, and identified that high-quality ecological areas are concentrated along the northeast-southwest transect, where they make up 20% of the landscape.

Why it matters — It introduces a direction-specific Ecological Quality Index that captures spatial inequities and ecological variations along specific geographic axes rather than assuming uniform radial urban decay. This provides planners in arid regions with a replicable method to pinpoint where agricultural fragmentation is most severe and where green infrastructure corridors are most needed.

Caveat: The empirical findings and the specific directional dynamics of the ecological index are demonstrated only for a single arid-region city, which may limit direct generalization to humid or monocentric urban forms.

Journal of Urban Management · pdf · doi

PedNStream: Scalable Network Flow Simulation for Pedestrian Traffic Management

The researchers developed PedNStream, an open-source, Python-native simulator that models macroscopic pedestrian network flows using the Link Transmission Model. The framework incorporates stochastic link dynamics to capture local flow variations, a utility-based route-choice model for adaptive rerouting under congestion, and controller interfaces for gating, flow separation, and route guidance. Its performance was validated through synthetic scenarios, real-network experiments comparing simulated flows to observed pedestrian counts, and a closed-loop control case study.

Why it matters — It provides a computationally efficient, scalable alternative to microscopic pedestrian models, allowing planners to repeatedly simulate and test real-time crowd management interventions across large-scale urban networks.

arXiv · pdf

Beyond the Flow: A Bayesian Latent Clustering Framework for Shared Micro-mobility Users in Venice

The study introduces a Bayesian finite mixture model for multivariate categorical count data that clusters micro-mobility users directly from repeated trip-level observations, preserving individual travel behavior structure. The framework was tested using a one-year dataset of over 220,000 shared bike and e-bike trips made by more than 11,000 recurrent users in Venice, Italy. The model successfully identified eight distinct latent mobility profiles, including localized, commuter-oriented, tourist-oriented, central, and inter-zonal travel behaviors.

Why it matters — By clustering users directly from individual trip records rather than aggregating data into user-specific summaries first, this method prevents ecological fallacies and preserves trip-level variability. This provides a scalable way to segment transportation users while accounting for uncertainty in cluster assignments.

arXiv · pdf

How Environment and Urbanization Shape Bird Diversity in Sri Lanka

This study analyzed bird diversity across Sri Lanka by combining observation records with environmental datasets, including weather, air pollution, NDVI, elevation, and Artificial Light At Night (ALAN). Spatial analyses were conducted across 2 km, 5 km, and 10 km grid scales using Poisson Generalized Linear Models and effort-corrected metrics like rarefied richness and occupancy rates. The analysis revealed that land-cover type is a stronger predictor of bird diversity than individual continuous variables, and that urbanization (measured via ALAN) reduces overall species richness while boosting a small number of generalist species.

Why it matters — It establishes a scalable, reproducible framework for tracking avian biodiversity under urban pressure, demonstrating that broad land-cover categories are more reliable indicators of species richness than isolated environmental metrics like temperature or greenness.

arXiv · pdf

Economic Disparities and Their Relationship to Destructive Health Behaviors in Five Western U.S. States

This study analyzed the relationships between county-level economic variables and adverse health outcomes, specifically suicide rates, across counties in Washington, Idaho, Oregon, California, and Nevada. Using data from the US Census and County Health Rankings, the analysis employed principal component analysis, clustering, correlation analysis, linear fittings, and LASSO regression after performing regression imputation on missing values.

Why it matters — The analysis demonstrates that county-level health and economic indicators tend to cluster along state-level boundaries, suggesting that state-level policy and economic environments play a significant role in local public health outcomes.

Caveat: The study relies on correlational methods and exploratory data analysis, meaning the identified relationships between economic variables and suicide rates cannot be interpreted as causal.

arXiv · pdf